Exp Brain Res (2002) 145:166–176 DOI 10.1007/s00221-002-1090-0
R E S E A R C H A RT I C L E
Margret Hund-Georgiadis · Ulrike Lex Angela D. Friederici · D. Yves von Cramon
Non-invasive regime for language lateralization in rightand left-handers by means of functional MRI and dichotic listening Received: 11 April 2001 / Accepted: 26 February 2002 / Published online: 8 May 2002 © Springer-Verlag 2002
Abstract Language lateralization was assessed by two independent functional techniques, fMRI and a dichotic listening test (DLT), in an attempt to establish a reliable and non-invasive protocol of dominance determination. This should particularly address the high intraindividual variability of language lateralization and allow decisionmaking in individual cases. Functional MRI of word classification tasks showed robust language lateralization in 17 right-handers and 17 left-handers in terms of activation in the inferior frontal gyrus. The DLT was introduced as a complementary tool to MR mapping for language dominance assessment, providing information on perceptual language processing located in superior temporal cortices. The overall agreement of lateralization assessment between the two techniques was 97.1%. Conflicting results were found in one subject, and diverging indices in ten further subjects. Increasing age, non-familial sinistrality, and a non-dominant writing hand were identified as the main factors explaining the observed mismatch between the two techniques. This finding stresses the concept of an intrahemispheric distribution of language function that is obviously associated with certain behavioral characteristics. Keywords fMRI · Dichotic listening · Language lateralization · Left-handedness · Age effects · Gender
Introduction The determination of language dominance has become an important issue in the presurgical benefit-risk evaluation of patients with brain tumors in eloquent areas and/ or intractable epilepsy (Loring et al. 1990; Gerschlager et al. 1998; Hajek et al. 1998). An adequate evaluation demands techniques that are reliable, reproducible, reM. Hund-Georgiadis (✉) · U. Lex · A.D. Friederici D.Y. von Cramon Max-Planck-Institute of Cognitive Neuroscience, Stephanstrasse 1, 04103 Leipzig, Germany e-mail:
[email protected]
peatable, and risk-free in nature, i.e., in the best case non-invasive. So far, the intracarotid amobarbital procedure (IAP; Wada and Rasmussen 1960) has been considered as the gold standard to assess language dominance, however, with major drawbacks. Its invasiveness provides an additional risk for the patient. Moreover, IAP cannot easily be repeated and the examination time is limited to the time frame of 20–30 min, therefore only restricted aspects of language organization can be addressed. Additionally, the spatial resolution of IAP is very poor. As the technique is based on behavioral testing of certain language functions during anesthesia of one hemisphere, it can solely characterize language subfunctions within the whole hemisphere. Although the procedure was initially used to lateralize language function, it has further evolved to include prediction of degrees of verbal memory decline following left temporal lobectomy. The reliability of IAP has recently been questioned, when language lateralization is solely and primarily based on aspects of speech production such as speech arrest (Benbadis et al. 1998). However, IAP procedure focusing on language comprehension as described by Loring and coworkers (1990) is able to assess language lateralization as a more fine grained variable rather than a discrete category, such as ‘left’, ‘right’, or ‘mixed’. The individual lateralization scores indicate a relative interhemispheric distribution of language functions with different contributions from various eloquent areas, although, without revealing their anatomical localization within the language network. Two main questions arise. First, what is the contribution of different eloquent cortices of both hemispheres to the processing of language? The answer is of interest in itself and leads straightforwardly to the second set of questions, namely, what are the clinical implications and applications of this line of inquiry? If language lateralization is a rather distributed effect, how does this knowledge affect surgical decisionmaking in individual patients and how does it influence reorganization of language function following stroke or hemorrhage in eloquent cortices? What is the clinical
167
outcome of patients with mixed dominance? Thus it appears that there are many reasons to thoroughly examine the concept of distributed language dominance in normal right- and left-handed controls rather than applying it to patients. Due to methodological and technical limitations, IAP cannot sufficiently address this new research line. A major effort has been made in precedent functional imaging research to replace the IAP by non-invasive fMRI mapping. Several comparative studies showed a reasonable agreement of the two techniques (Demb et al. 1995; Desmond et al. 1995; Binder et al. 1996). However, sole reliance on the alternative technique for identification of dominant eloquent cortices does not seem to be considered routinely as adequate and sufficient in the clinical setting. There are also some caveats in using fMRI-assessed lateralization data as a basis for surgery. First, lateralization is purely based on activated pixels in predefined eloquent cortices and does not necessarily relate directly to behavioral data. Second, to isolate the relevant cognitive process, the MR design is based on a comparison of the relevant target task with an appropriate baseline condition. Even slight changes of the experimental protocol may significantly influence the outcome, i.e., the lateralization index, and thereby may complicate clear functional decision-making. Third, almost all fMRI lateralization studies report contributions of the subdominant hemisphere, that have not been expected to this extent on the basis of either IAP and lesion studies. The role of the subdominant cortices in the overall language network identified by fMRI studies has to be further evaluated in future studies, beyond the sole determination of laterality. Answers to these issues are needed before the role of fMRI in the lateralization of language function can be fully determined. Previous fMRI research has applied various paradigms to assess the dominant language site of the brain, in particular the fronto-opercular cortices. Much weight has been put on finding both the relevant target and the appropriate baseline task controlling for non-relevant cognitive processes. Word and sentence production tasks (Hertz-Pannier et al. 1997) as well as word classification tasks (Desmond et al. 1995; Gabrieli et al. 1995; Binder et al. 1996) have been used for this purpose. The activation in the inferior prefrontal gyrus (BA45, 46, 47) during lexical and semantic encoding tasks has been claimed to define global hemispheric dominance of speech. It is intriguing that the choice of the cut-offs for the categories ‘left’, ‘right’, and ‘bilateral’ was found to influence the statistical distribution of lateralization. Moreover, different genetic, developmental, and environmental factors may influence the functional organization patterns of language as assessed by means of fMRI. So far, there are only vague ideas about the influence of age (Ross et al. 1997; D’Esposito et al. 1999), as most functional mapping studies are based on young control subjects. It has been claimed that a change of the hemodynamic coupling of the blood oxygenation level dependent (BOLD) to the neural response may accompany ag-
ing. Moreover, there are contradictory statements about the influence of sex (Shaywitz et al. 1995; Frost et al. 1999; Springer et al. 1999). While Shaywitz et al. (1995) described a more bilateral dominance distribution in female healthy controls, Springer et al. (1999) and Frost et al. (1999) did not find any significant sex differences and argued against substantive differences between men and women in the large-scale neural organization of language processes. Although the influence of hand dominance on the cerebral organization of language has been widely examined, it remains poorly understood. The influence of certain cofactors of left-handedness, such as familial sinistrality, inverted writing, or ambidextrous writing on the cerebral organization patterns is unclear. No doubt, these characteristics may contribute to a large intersubject variability, thereby complicating functional decisionmaking. If functional mapping research aims at routinely replacing IAP, a clear regime is required to deal with intraindividual inconsistencies and deviations. Individual variability seems to pose an intriguing problem. Within the ever-expanding language mapping studies, fMRI data of normal control cohorts were recently collected (Springer et al. 1999). They reflect the expected distribution of language dominance in normal controls and show reasonable agreement with IAP. The single case though, especially when pathological brain structures are involved, may sometimes allow ambiguous and contradictory interpretations. Hence, definite determinations based on fMRI would profit from a ‘second opinion’, which should be acquired by other non-invasive behavioral tests or an extended MR protocol. In the years since Kimura’s initial reports (1967), the dichotic language test (DLT) has gained wide usage in both clinical and experimental settings. The direction of ear advantage (left versus right) is generally accepted as an indicator of the hemisphere which is superior for speech discrimination. The DLT thus enlightens a restricted aspect of language processing. The auditory discrimination at the word level is supposed to engage components of the language network related to perception and has been demonstrated to activate the superior temporal gyrus of both hemispheres to a different extent (Lee et al. 1994; Hugdahl et al. 1999). In contrast, the currently used fMRI lateralization paradigms are designed to activate inferior frontal cortices, which are more dedicated to language production. Comparison between DLT- and IAP-based language lateralization disclosed a high correlation in both the overall and the inbetween test results (Strauss et al. 1987; Zatorre 1989). In the present study, language lateralization was assessed by both fMRI and DLT to evaluate the agreement of these non-invasive techniques. The issue of intersubject variability was particularly addressed, seeking to provide a diagnostic strategy appropriate to functional decision-making in individual subjects. It aims at characterizing the functional organization patterns of a relevant aspect of language in a group of left-handed subjects (LH) in comparison to right-handed subjects (RH), tak-
168
ing into account the influences of sex, age, hand dominance, and familial sinistrality.
Materials and methods Thirty-four healthy control subjects (18 male, 16 female), 17 RH and 17 LH, aged from 20 to 67 years (32.4±15.87) were included in the study. Eight out of 17 LH had a history of familial sinistrality. Nine subjects were ambidextrous. All participants had normal or corrected to normal vision. To examine age-dependent lateralization effects, 10 subjects (5 male, 5 female) aged over 40 years were included in the study. The group statistics was performed for the entire groups of RH and LH, in a second step for the two age subgroups (G1, age 40 years), and in a third step separately for males and females. The experiments were generally approved by the local ethics committee, and all subjects gave written informed consent to participate in the study. Behavioral tasks Each subject performed the Edinburgh handedness inventory (EHI) to assess the degree of handedness (Oldfield 1971). Speech lateralization was determined in behavioral terms using an auditory dichotic listening test based on the discrimination of rhyme words (DLT; Diesch and Haettig 1997). In the task ten monosyllabic pairs of rhyme words were simultaneously presented to the left and the right ear. The subjects were asked to loudly repeat the word which they had identified. A lateralization index (LI) was calculated from the number of detected right (nPr) and left ear stimuli (nPl) [LI=nPr-nPl/nPr+nPl (range±1)]. Positive values indicate right ear and left-hemispheric dominance. Bilaterality was defined when values ranged between ±0.1. The DLT is based on the principle that, whenever two auditory signals are put in conflict, the contralateral input is suppressing the ipsilateral signal. Therefore, the ear contralateral to the dominant language cortex should show an advantage over the ipsilateral ear. Activation paradigms During the fMRI experiment, common German words composed of two syllables (length: four to seven letters) were visually presented every 2 s in a word classification task. The target stimuli consisted of nouns and verbs, half abstract and half concrete, divided in eight blocks of 20 words each, with a pseudorandom presentation order. Semantic (abstract or concrete) encoding was compared to a perceptual task (presentation of words with normal or double spacing of letters). The lexical encoding task (noun or verb) was contrasted with a different perception task (presentation of words in capital or small letters). The baseline condition was either one of the two perceptual encoding conditions or a resting condition (dark desktop consisting of five white Xs presented for 40 s). The order of the MR scanning protocol varied between subjects and started randomly with the lexical encoding or with the semantic encoding tasks. The baseline condition task, i.e., perceptual encoding versus resting condition constituted the last part of the scanning protocol. All applied word stimuli were presented only once within the overall protocol to counteract effects of task priming. During the fMRI experiment behavioral data (reaction times and number of errors) were recorded online.
Image analysis Functional MRI data were analyzed with the BRIAN Software (Kruggel and Lohmann 1996). The initial four words target items were excluded from analysis to avoid influence of vascular arousal caused by the onset of the scanner noise. Preprocessing included movement correction in 2D, and baseline correction entailing filtering of functional data in the temporal domain by a lowpass filter. Functional images were created by generating statistical z-maps with z-values >5 on a single pixel level. The regions of interest were characterized in size, volume, and intensity of activity. Lateralization indices were calculated by the number of activated pixels (P) with a z-score higher than 5.5 in Broca’s area [Talairach and Tournoux (1988): cC6, cC7, cD5, cD6, cD7], in the superior temporal gyrus (Talairach: cF10, cF9, cG10, cG9, cE9, cE10), and for each hemisphere [LI=nPl-nPr/nPl+nPr (range±1)]. Lateralization indices were subsequently classified according to Strauss et al. (1987) as left hemisphere language dominant (defined as LI>0.1), symmetric (–0.1≤LI≤0.1), and right hemisphere dominant (defined as LI3) separately listed for right-handers (RH) and left-handers (LH) during semantic versus perceptual encoding. (GFI Inferior frontal gyrus, GFM medial frontal gyrus) Talairach coordinates Volumes (z-max) Area
Left
Right
RH (n=17) GFI (BA44) Anterior insula
–42 11 27 – – – –37 20 6 – – –
Total volume (mm3) LH (n=17) GFI (BA44) GFI (BA45) GFM (BA46/9) Anterior insula
Left
1,522 (3.98) – 164 (2.87) – 1,686
–36 1 30 –42 17 16 ––– –39 21 0
35 3 32 43 24 20 35 27 27 –––
Total volume (mm3)
Right
395 (2.90) 647 (2.82) – 30 (2.26) 1,072
– 208 (2.66) 132 (2.56) 131 (2.49) – 471
Table 2 Talairach coordinates and volumes (mm3) of the main activations in eloquent cortices (z>3) separately listed for right-handers and left-handers during lexical encoding versus perceptual encoding Talairach coordinates Volumes (z-max) Area
Left
Right
Left
RH (n=17) GFI (BA44/6) GFI (BA44/45) Anterior insula Thalamus
–47 6 16 –40 16 15 –26 8 6 –1 –7 17
––– ––– ––– –––
11 (1.59) 496 (2.34) 72 (1.78) 84 (1.97)
Total volume (mm3) LH (n=17) GFI (BA44/6) GFI (BA44/45) Anterior insula Thalamus Total volume (mm3)
Right
663 –44 5 15 –40 19 13 –30 20 –1 –7 –19 15
––– ––– ––– –––
9 (1.61) 120 (1.90) 10 (1.60) 84 (1.76)
Fig. 2 Different potential of the applied word classification and baseline tasks to show laterality. The mean laterality indices of the various tasks are displayed for the group of right-handers. Semantic encoding contrasted with a perceptual control task was the best indicator of lateralization
language, 1 showed symmetric language organization, and 7 were defined as right dominant related to cortical language representation (BI for the entire group 0.19±0.59). Language dominance could definitely be determined in 33 out of 34 examined subjects (97%) by means of fMRI. The more global hemispheric lateralization index (HI) yielded comparable results: HI was 0.52±0.34 and 0.14±0.6 in the RH and LH groups, respectively. A highly significantly correlation (r=0.9, P5.5) was assessed in 18 out of 34 subjects with mean TIs of 0.49±0.2 and 0.07±0.4 for RH and LH, respectively. Statistically significant correlations were also found between TI and BI (r=0.8, P
